To appear in Hohwy and Kallestrup, eds. Being Reduced. Oxford University Press.
Mental Causation and Neural Mechanisms * Jim Woodward
Caltech
Issues about the causal role of the mental – about whether mental states (beliefs, desires, intentions and so on) can cause or figure in causal explanations of other mental states or behavior, about what it even means to attribute causal efficacy to a mental state, and about how claims about mental causation/explanation fit with (or fail to fit with or are undermined by) claims about causation by neural mechanisms – have been matters of intense debate within the philosophical literature over the past decade. Some philosophers argue that generally accepted claims about what makes a relationship causal and about the relationship between mind and body yield the conclusion that mental states cannot cause anything -- that the mental is entirely causally inert or epiphenomenal. The arguments for this conclusion are largely metaphysical and quasi-apriori in the sense that the conclusion is supposed to follow from the combination of very general and presumably uncontroversial empirical assumptions about the relationships between the mental and the physical (the causal closure of physics and the absence of systematic causal over-determination of mental states by both mental and physical causes) together with assumptions about what is involved in mental causation. Other philosophers have found this conclusion literally incredible and have sought to identify flaws in the
arguments that seem to support it. However, no particular counterargument has won general acceptance.
In this paper, I propose to examine these issues within the framework of the account of causation and causal explanation worked out in my recent book, Making
Things Happen (MTH). One of my themes will be that many of the standard arguments
for the causal inertness of the mental rest on mistaken assumptions about what it is for a relationship to be causal, and about what is involved in providing a causal explanation. These mistaken assumptions involve an inter-related complex of ideas, described below: a conception of causation according to which a cause is simply a condition (or a conjunct in a condition) which is nomologically sufficient for its effect, and the closely associated deductive-nomological (DN) conception of explanation according to which explaining an outcome is simply a matter of exhibiting a nomologically sufficient condition for it. Given these assumptions, it is indeed hard to understand how there can be such a thing as mental causation. However, the account of causation defended in MTH undercuts these assumptions and in doing so, allows us to reach a better understanding of what is involved in mental causation and of the real empirical issues surrounding this notion.
1.
MTH defends a manipulationist or interventionist account of causation: causal (as opposed to merely correlational) relationships are relationships that are potentially exploitable for purposes of manipulation and control. As an illustration of what this means, consider the well known correlation between attendance at a private (that is, non government run) secondary school in the contemporary U. S. and scholastic achievement: students who attend private schools tend to score higher on various measures of
scholastic achievement than students who attend public schools. This correlation raises the question of whether private school attendance causes superior scholastic achievement or whether instead the relationship between these two variables is merely correlational, with the correlation between them due to the causal influence of some other variable(s). To take only the most obvious possibilities, it may be that parents with higher SES are more likely to send their children to private schools and that SES (socio-economic status) also directly causes scholastic achievement. Or it may be that parents who send their children to private schools tend to value educational achievement more and these values directly influence their children’s performance. If we let P be a variable measuring whether a child attends public or private school, S a variable measuring scholastic achievement, and E and A be variables measuring, respectively, parents’ social
economic status and attitudes toward education, these possibilities might be represented as follows, with an arrow from X to Y meaning that X causes Y:
E
P S
A
P S
On a manipulationist conception of cause, the question of whether P causes S is identified with the question of whether S would change under some suitable manipulation of P. If P causes S, then other things being equal, this will be a good or effective strategy. If on the other hand, if P and S are merely correlated as in Figure 1, changing the school the child attends should have no effect on achievement. Instead changing SES or parental attitudes would an effective strategy for affecting achievement.
How might one determine whether S would change under a suitable manipulation of P and what does “suitable” mean in this context? One possibility would be to perform a randomized experiment: children in the population of interest are randomly assigned to one of two groups, one of which is sent to private schools and the other to public schools. One then looks to see whether there is a correlation between P and S. The effect of the randomization (it is assumed) is to remove any systematic difference between the two groups with respect to parental SES, attitudes, or indeed any other factors that might influence S independently of P. Any remaining correlation between P and S should thus be attributable to the causal influence of P on S. If Figure 1 represents the correct causal structure there should be no correlation between P and S under any such intervention on
P.
A natural way of representing such a randomized experiment, due to Spirtes, Glymour and Scheines, 2000 and Pearl, 2000 is to think of the experimental manipulation of P (represented by means of a variable I for intervention) as accomplishing the
Figure 1.2
On the other hand if, say, the correct causal structure is one in which it is true
both that E is a common cause of P and Sand that P causes S (i.e. the correlation between P and S is due to both of these factors) then the result of intervening on P is to replace the structure
with the structure:
I P
E
S
I P S
I P S
E
Figure 1.3
In this case there will be a change in S under an intervention on P, reflecting the fact that (unlike the situation represented in Figure 1) P makes a causal contribution to
S that is independent of, or in addition to, the contribution made by E.
Note that if we want to do an experiment of this sort to determine whether P
causes S it is crucial to the logic of the experiment that the intervention not itself cause or be correlated with other causes of S that are independent of P. For example, if Figure 1a is the correct structure, an alternative way of manipulating P (besides what is represented by Figure 2) would be to manipulate E (perhaps we give parents a very large cash grant). This manipulation of E would change the value of P in the population (since E causes P), but it would (obviously) not be a good experimental design for determining whether P
causes S since it confounds any effect of P on S with the effect of changing E on S. Instead, what we want is that, among other desiderata, the experimental manipulation be such that the variation in P it introduces is uncorrelated with or independent of other possible causes of its putative effect S (except of course for those other possible causes that lie on any causal route (should one exist) from P to S. An experimental manipulation of P that has this feature and also features that rule out other confounding possibilities is what we mean by an intervention.
variables supervene on others, the issue of what counts as such a confounding possibility requires some careful thought—this is addressed below, especially in Section 6.
Given this notion, we may use it to give an interventionist characterization of what it is for a variable X to cause or be causally relevant to a second variable Y. (I will use “cause” and “causally relevant” interchangeably and in a generic sense according to which X causes Y if it is either positively or negatively relevant or of mixed relevance for
Y).
(M) X causes Y if and only if there are background circumstances B such that if some (single) intervention that changes the value of X (and no other variable) were to occur in B, then Y would change1.
(M) obviously requires some explication. First, note that it relates variables, which as Woodward, 2003 explains, are the natural candidates for the relata of causal claims within an interventionist framework. A variable is simply a property, quantity etc, which is capable of taking two or more “values”. Philosophers often focus on causal claims relating types of events, and we can think of these relata as two-valued, with the values in question corresponding to the presence or absence of this type of event. For example, we may think of the claim that short circuits cause fires as relating variables which take values corresponding to <short circuit present, short circuit absent>, and <fire present, fire absent>. However, some variables such as pressure or mass may take many different values.
The reference to background conditions is added to accommodate the familiar fact that it may be that it is only under certain conditions, not specified in the description of X
itself, under which interventions on X are associated with changes in Y. Thus, for example, according to M, short circuits cause fires as long as it is true that in some background circumstances (having to do with the presence of oxygen etc.) interventions that change whether a short circuit occurs are associated with changes in whether a fire occurs (or in the probability of fire).
Next, note that the formulation M relates changes in X (due to an intervention) to changes in Y (or in the probability distribution of Y). Focusing first on the case in which the causal claim relates changes in the value of X to the changes in the value of Y, I take this to imply that there is a pattern of association between X and Y such that each of these variables can take at least two different values (X=x, x’ with x x’, Y= y, y’ with y
y’ ) such that one (e.g., x) of these values of X (when produced by an intervention) is associated with one (y) of the values of Y and a different value x’ of X (when produced by an intervention) is associated with a different value y’ of Y. That is, X causes Y if and only if there are distinct values of X and Y meeting the conditions just described and background circumstances B in which two counterfactuals of the following form are true:
(M*)
(M1*) If an intervention that sets X=x were to occur in B, then Y=y. (M2*) If an intervention that sets X=x’ were to occur in B, then Y=y’.
When M1* and M2* hold, I will say that a change in the value of X from X=x to
X=x’ (where x x’) in background circumstances Bcauses a change in the value of Y
from Y= y to Y=y’(and vice-versa).
intended as an explication of the notion of cause that figures in claims like “attendance at private school causes improved scholastic achievement” (alternatively: “a change in attendance from public to private school causes a change in scholastic achievement from better to worse”) “smoking causes lung cancer” as opposed to such token claims as “Smith’s attendance at private school in 1990 caused his scholastic achievement in the same year to improve” or “Jones’ smoking caused his lung cancer” . As MTH shows, the interventionist account can also be used to capture a notion of token causation, but with the exception of some remarks about pre-emption and redundancy in section 6, my focus in this essay will be entirely on type causal notions of the sort captured by M and on type causal claims about mental causation. The reason for this focus is that I take issues about the causal role of the mental to be in the first instance issues about type casual claims involving mental states – whether beliefs, desires intentions cause other mental states or behavior. If such claims about mental causation are never true, then presumably it is also never true that, e.g., some particular token mental state of Jones caused some bit of his behavior. The latter token claims also, however, raise some distinctive issues of their own that for the purposes of this essay are simply distractions.
As the previous paragraph makes explicit, (M) embodies a counterfactual account of causation in the sense that it links the claim that X causes Y to a claim about what would happen to Y if, perhaps contrary to actual fact, an intervention on X were to occur – what I will call an interventionist counterfactual. As MTH explains in more detail, the conditions that characterize the notion of an intervention do roughly the same work as the similarity metric in Lewis’ version of a counterfactual theory of causation: given an appropriately characterized notion of an intervention, the counterfactuals that figure in M will be non-backtracking, the joint effects of a common cause will not be counterfactually dependent on one another when dependence is understood in terms of interventionist counterfactuals, and other standard counter-examples to counterfactual accounts of causation will be blocked.
I assume that interventionist counterfactuals and the causal claims associated with them can be true even if the interventions that figure in their antecedents cannot in fact be carried out by human beings because of practical or other sorts of limitations. However, I also assume that if a candidate causal claim is associated with interventions that are impossible for (or lack any clear sense because of) logical, conceptual or perhaps metaphysical reasons, then that causal claim is itself illegitimate or ill-defined. In other words, I take it to be an implication of M that a legitimate causal claim should have an intelligible interpretation in terms of counterfactuals the antecedents of which are coherent or make sense.
As an illustration, the claim that an asteroid impact caused the extinction of the dinosaurs can be understood within an interventionist framework as a claim about what would have happened to the dinosaurs if an intervention had occurred to prevent such an asteroid impact during the relevant time period. In this case we have both (i) a
occur. By contrast, neither (i) nor (ii) hold if we are asked to consider hypothetical interventions that make it the case that 2+2 4 or that the same object is at the same time both pure gold and pure aluminum or that transform human beings into houseflies. Causal claims that require for their explication claims about what would happen under such interventions (“2+2 = 4 causes it to be the case that…”) are thus unclear or at least have no legitimate role in empirical inquiry. This idea – that the counterfactuals that are relevant to the explication of causal claims must have a clear interventionist
interpretation—will play an important role below.
A closely related idea, to which I will also appeal, is that genuinely competing or rival causal claims must make different predictions about what would happen under some possible intervention or interventions, where the interventions in question are again such that we have some coherent conception of what it would be like for them to occur. Thus if we have two apparently competing claims, the first contending some mental state is causally inert and the other contending that it causes some outcome, it must be possible to specify some set of (coherent, well-defined) interventions such that the two claims make competing predictions about what would happen under those interventions. If we cannot associate such an interventionist interpretation with one or both of the claims, the claim(s) in question lack a clear sense and if they fail to make different predictions about what would happen under such interventions, they are not genuine competitors.
Note that M characterizes a rather weak and non-specific notion of “cause” : for
changes in X are associated with such changes – see example (1.1) below. Ideally, of course, we would like to know much more than this: we’d like to know exactly which changes in X are associated with exactly which changes in Y and in what background circumstances. Within the interventionist account such information is spelled out in terms of more fine-grained interventionist counterfactuals specifying in a detailed way how Y changes under various interventions that change the value of X.
Next, let me make explicit a consequence of the fact that (M) relates changesin the value of X to changes in the value of Y: as the formulation in terms of M*
underscores, (M) implies that for X to count as a cause of Y, there must be at least two different values of X, x and x’, and at least two different values of Y, y and y’ such that under an intervention that changes the value of X from x to x’, the value of Y changes from y to y’. This captures the intuition that whatever else a cause is, it must make a difference to its effect: to say that a cause makes a difference to its effect implies that there must be at least two possible states, corresponding to the cause being present or the cause being absent or to the cause being in one state rather than another state such that which of these is realized is associated with some change or difference in the state of the effect (the difference between the effect being present or absent etc.)2 Causal claims thus have built into them the feature that philosophers call contrastive focus or “rather than” structure: when we make the structure of a causal claim explicit, we see that the real content of the claim is something like this: it is the contrast between X’s taking some value x and its taking some different value x’ that causes the contrast between Y’s taking
2 I have glossed over a number of issues here. For a discussion of how this difference-making conception of causation can accommodate examples involving
over-determination etc. see Woodward, 2003, Chapter 2. The idea that causes must be difference-makers does not by itself exclude the possibility that causes must also satisfy other conditions that can’t be captured in difference-making terms, such as those
value y and value y’ (or alternatively it is the fact that X=x rather than X=x’ which causes Y to be y rather than y’ or the change from X= x to X=x’ that causes the change from Y= y to Y=y’) The causal claim that it is the contrast between X=x rather than X= x’
that causes the contrast between Y= y rather than Y=y’ is thus a different causal claim, with a different content, from a causal claim involving a different contrast in the values of X such as the claim that X=x rather than X= x’’ (where x’’ x’) accounts for the contrast between Y= y rather than Y=y’’.
We noted above that there are many cases in which some changes in the value of a candidate cause variable X will be associated with changes in the value of a candidate effect variable Y , but other changes in which X will not be associated with changes in Y. Making the contrastive focus of a causal claim explicit is a natural way of representing such facts. When the contrastive focus of a causal claim is not made explicit, there may or may not be a natural default specification of the contrast situation which corresponds to the cause being different or absent. As an illustration of the first possibility, if the claim of interest is that short circuits cause fires, the natural default contrast (if this is not explicitly specified) is a situation in which no short circuits of any kind and no
state and different outcomes associated with each. In such cases, the causal claim may be ambiguous unless we make clear what contrastive focus is intended.
Whether or not there is a natural default, a causal claim will be defective to the extent that it suggests that some contrast or difference in the value of the cause variable is associated with changes in the effect variable when this is not the case or if it fails to make explicit which changes in the cause variable are associated with which changes in the effect variable,( as would be the case, for example, if there is no obvious default and the contrastive focus is not specified). In some cases of this sort, it may seem most natural to think of this defect as a matter of the causal claim being false and in other cases, more natural to think of the claim as true but misleading or as failing to covey information that it should convey. For our purposes, it often will not matter much which of these alternative assessments is adopted.
As an illustration, consider a platform that will collapse if and only if a weight greater than 1000 kg is placed on it. If it is claimed that (1. 1) it is the fact that the weight on the platform was greater than 1000kg that causes the platform to collapse, this is naturally interpreted as the claim that it is the contrast between the weight being greater than rather than less than 1000kg that caused the platform to collapse – a claim that is correct in the specified circumstances. That is, the weight’s being less than 1000kg is the natural default for the absence of the cause when no explicit contrast is specified.
comparison with (1.2), in failing to communicate information about the conditions under which the platform would not have collapsed. Put in terms of M*, (1.2) does not tell us which changes in the weight cause changes in whether the platform collapses (or not) . This observation will turn out to be important in connection with claims about mental causation.
Finally, note that according to (M), if no changes (produced by interventions) in the value of X are associated with changes in the value of Y , then X does not cause Y. Instead, X is causally irrelevant or causally inert with respect to Y. Put slightly
differently, if we understand causal (ir)relevance in the manner just suggested ( X is causally relevant to Y if and only if there is at least one change in the value of X such that if it were produced by an intervention, there would be a change in the value of Y), there is no such thing as a cause of Y that is not causally relevant to Y. Equally, if X is causally relevant to Y, then X causes Y. Bona fide causal claims always have relevance claims built into them. I stress this point because some influential writers on mental causation seem to assume (more or less explicitly) that there is a notion of causation or causal efficacy according to which X can cause Y without being causally relevant (in the sense just defined) to Y or, alternatively, that X can be causally relevant to Y, without its being true that X causes Y.
the point of view of explanation – superior because they are, e.g. more general or provide more information about the conditions under which alternatives to the
explanandum outcome occur. (See below for illustrations) But these more general etc. causal claims are still just ordinary causal claims that must (if interventionism is correct) possess the sorts of features set out in M – they are not some special sort of causal claim with special features that play a role in causal explanation but not in other kinds of causal ascription. In my view, there is thus a connection between features of our explanatory practice involving mental events and the truth of causal claims about the mental in the following sense: if various features of our practice of giving causal explanations involving mental events are correct or well-founded, then the causal claims figuring in those explanations must be true.
With this as background, let me flesh out the interventionist conception of causal explanation a bit: we may think of this as embodying a what if –things had been -different conception of explanation: we explain an outcome by identifying conditions under which the explanandum-outcome would have been different, that is, information about changes that might be used to manipulate or control the outcome. More generally, successful causal explanation consists in the exhibition of patterns of dependency (as expressed by interventionist counterfactuals) between the factors cited in the explanans
and the explanandum – factors that are such that changes in them produced by
in the explanandum – phenomenon (such factors are causally or explanatorily irrelevant to the explandandum phenomenon).
How does this conception of causal explanation compare with the well-known deductive –nomological (DN) model of explanation, according to which we explain an explanandum by deriving it from a “law” and other true statements (typically about “initial conditions) and in this sense exhibiting a nomologically sufficient condition for it? One crucial difference3, which is of central importance in the mental
causation/explanation debate, is that the DN model does not impose the requirement that a successful explanation answer a what- if things –had- been- different question. This is the source of a number of well-known counterexamples to the DN model. Consider the following derivations, due to Wesley Salmon (1984):
All men who take birth control pills fail to get pregnant. Jones is a man who takes birth control pills
Jones fails to get pregnant
All samples of hexed salt dissolve in water This is a sample of hexed salt
This dissolves in water.
In both cases, the derivations are sound and the generalizations in them satisfy the criteria for lawfulness found in the philosophical literature. Nonetheless, the derivations don’t seem explanatory. In both cases, the underlying defect seems the same: the derivations cite conditions that, although nomologically sufficient for their explananda, are not
causally relevant (in the sense captured by (M)) to those explananda. That is, we judge that the above explanations are defective because they cite conditions that are not causes of (or causally relevant to) the outcomes they purport to explain. For example, changes in whether Jones takes birth control pills (when produced by interventions) are not associated with changes in whether he gets pregnant, and, in accordance with (M) , this is reflected in our judgment that taking birth control pills does not cause and is not causally relevant to Jones’ failure to get pregnant. Similarly, changes in whether the salt is hexed or not (when produced by interventions) are not associated with changes in whether it dissolves and this is what accounts for our judgment that the hexing does not cause and is not causally relevant to the dissolving. The causal irrelevance of Jones’ taking birth control pills and the hexing of the salt to these outcomes is also reflected in the fact that explanations that appeal to these factors do not provide answers to what- if-things-had-been-different questions about these outcomes. As these examples illustrate, citing a nomologically sufficient condition for some outcome is not the same thing as answering a what-if things-had-been-different question with respect to that outcome. Similarly, contrary to what a number of philosophers of mind seem to suppose, a
condition that is linked by law to an outcome is not necessarily a condition which causes or is causally relevant (in the sense of cause and causal relevance captured by (M)) to that outcome.
A parallel observation applies to accounts of causation which take C to be a cause of E if and only if C is a nomologically sufficient condition for E (or what we might loosely call a “part” or “conjunct” in such a condition). The hexing of the salt is
judgments: the conditions cited as causes are not such that any interventions on them are associated with changes in their putative effects.
Let me conclude this section with some brief remarks about an issue that will be a source of concern to a number of readers. The account sketched above links causal claims and explanations to the truth of certain interventionist counterfactuals. A common
contention is that counterfactuals cannot be “barely true” – instead they require “truth makers” that presumably must be specified in non- counterfactual terms. “Laws” are the usual candidates for such truth-makers. For a variety of reasons I am skeptical about this contention, but, as nearly as I can see, nothing will turn in what follows on what we take the truth makers for counterfactuals to be. What matters for the arguments that follow is whether causal claims and explanations are related to interventionist counterfactuals in the way that I have claimed—any account of the truth conditions for counterfactuals that is consistent with these relationships will be acceptable for the purposes of this essay.
That having been said, there is a certain tempting but plainly mistaken inference that we need to be careful to avoid. The inference goes something like this:
counterfactuals require laws as truth makers; therefore, any account of causation in terms of counterfactuals is committed to an account according to which all that is involved in one item, property etc. A causing or being causally relevant to another B is that A be linked by law to B. In other words, the inference is from the claim that the laws are the truth makers for counterfactuals (and causal claims) to the conclusion that a nomological sufficiency account of causation and causal relevance is adequate.
specifying that B would be different or be absent under some intervention that changes
A or causes it to be absent must be true. The fact – if it is a fact -- that there is some law (L) specifying that A or A in conjunction with other factors K is nomologically sufficient for B, does not settle the question of whether the counterfactual (F ) is true: As described above, (L) is not in itself a truth maker for the counterfactual (F), even though (L) is a truth maker for other counterfactuals that matter for the truth of (C) such as (F*) “If A
were to occur as a result of an intervention in K, then B would occur”. Of course if the contention about counterfactuals requiring laws as their truth makers is correct, then (F) will also have some law as its truth maker but this observation does not undermine the claim that (F) and not just (F*) is relevant to the truth of the claim that As cause Bs.
2.
What are the implications of the framework described in Section 1 for the status of “mental causation”? Prima-facie, it seems to support the claim that mental states can be causes. We do after all seem to regularly (and successfully) intervene to change the mental states of others and perhaps our own mental states as well and these changes in turn sometimes seem to be regularly associated with changes in other mental states and in behavior. Indeed, this seems to be what successful persuasion and deception are all about -- in persuasion I manipulate your beliefs and desires by providing you with
information or material inducements, typically with the goal in mind that these changes will in turn lead to further changes that I desire in your mental state or behavior. On an interventionist conception of cause, this is all that is required for mental causation – nothing more metaphysically portentous is needed. That is,
M2 (or B) will change. Common sense certainly supposes that episodes like these are very widespread.
Moreover, mental causation in the interventionist sense doesn’t seem confined to such contexts. Many experiments in psychology and the social sciences are naturally regarded as involving, among other things, successful attempts by the experimenters to manipulate subject’s beliefs by giving them verbal instructions (about e.g., what the experimental task is, what they will be rewarded for doing etc), where the goal of the experiment is to discover how these changes are systematically associated with changes in subjects’ behavior. Similarly, it is very natural to interpret many experiments (in, e.g. social psychology and experimental economics) involving interactions between people as investigations of, among other things, how changes in subject’s beliefs about one another’s beliefs and desires cause changes in behavior. For example, changes in my beliefs about how likely you are to cooperate in an iterated prisoner’s dilemma or trust game will cause changes in my behavior toward you, changes in responder’s beliefs about the alternatives available to the proposer in an ultimatum game will cause changes in the probability of responder rejecting the proposer’s offer and so on. Even
experimental demonstrations that show that certain beliefs do not, contrary to what subjects and others expect, causally influence subject’s behavior (as with experiments that show a position effect in the choice among identical consumer items and that subsequent reason giving is confabulation) seem to require some conception of what would be evidence for a causal influence of belief on behavior – it is the failure to find such evidence that shows the belief to have no causal influence. There would be no point in performing the experiment if beliefs could never, as a matter of principle, causally influence behavior.
Although the notion of mental causation thus seems, at least on the surface, unproblematic from a interventionist perspective, the philosophical literature is full of arguments to the contrary— arguments that purport to show that mental states or
properties cannot (ever) cause other mental states or behavior. In what follows I want to explore some of these arguments. I will begin at a relatively general and intuitive level and then consider some more precise arguments
One motivation for skepticism about assigning any causal role to the mental derives from the assumption that mental states are “multiply realizable”4 by different neural or physical states, combined with the thought that there is a general preference for detailed or fine grained or more micro level causal claims/ explanations (in this case claims at some physical or neural level) over less fine grained, more macro (e.g mental or psychological) claims. Suppose that my intention I to reach for a grape on some
particular occasion is followed by my reaching for that grape (Call this behavior R) Assume (as is standard) that I has neural/physical “realization” N1 (on this particular occasion), but that this same type of intention I might also have had a number of other possible realizations N2 ,N3, … etc. where the description of each of these realizations
4 The thesis that mental states are (or can be) multiply realized by different neural states is received wisdom in philosophy of mind and in part just for ease of exposition I will adopt this terminology in what follows. It is important to emphasize, however, that this thesis is unclear in a number of crucial respects, as recent discussion (e.g., Shapiro, 2000) has emphasized. I will also add that although it is often claimed that if mental states are multiply realizable, then this rules out the possibility of any theory according to which types of mental states or mental properties are identical with types of physical states or properties, this is simply a non-sequitur. As emphasized above, there is nothing in the idea of multiple realizability per se that rules out the possibility that all of the different realizers share some common physical structure at an abstract level of description. For example, different realizations of the same intention may share some aggregate feature that is a function of firing rates exhibited by a group of neurons, just as the same temperature may be realized by a variety of molecular configurations, all of which possess the same average kinetic energy. To the extent this is so, it may
contains a great deal more fine grained detailed information than the description that just adverts to I. If, furthermore, N1 is by itself (nomologically) sufficient for the occurrence of R, given the rest of the condition of my brain, why isn’t it at least preferable and perhaps mandatory to think of N1 as causing or causally explaining R? And once we do this, what causal or explanatory role can I play?
To explore the cogency of this reasoning, let us consider some other examples involving a choice between more or less fine-grained causal information.
3.1) Suppose that a mole of ideal gas at temperature T1 and pressure P1 at time
t1 is confined to a container of fixed volume V. The temperature of the gas is then increased to T2 by the application of a heat source and the gas is allowed to reach a new equilibrium at time t2 where its pressure is found to have increased to P2. One strategy (the macroscopic strategy) for explaining (or exhibiting the causes of) the new pressure is to appeal to the ideal gas law PV=nRT which describes the relationship between the macroscopic variables pressure, temperature and volume. According to this law, when the temperature is increased to T2 and the volume remains fixed, the new pressure at equilibrium must increase to P2= nRT2/V.
Now contrast this with the following, (entirely impractical) microscopic strategy for explaining the behavior of the gas: one notes the exact position and momentum of each of the 6x1023 molecules making up the gas. Call this configuration G1. From G1, the details of how the temperature source contributes to the kinetic energy of each of the individual molecules and knowledge of the exact laws governing the interactions between each of these molecules, one traces the trajectory of each gas molecule through time, eventually ending up with the exact position and momentum of each molecule making up the gas at time t2. This molecular configuration, G2, communicates a net force per unit area to the surface of the container which is just the new pressure P2.
molecules with the required exactness and the 6x1023 body problem of their interaction is of course computationally intractable But the inadequacies of this strategy do not just have to do with our epistemic limitations. There is a more fundamental difficulty: while the strategy succeeds in tracing the particular trajectories of individual molecules that in fact led on this particular occasion to the macroscopic outcome P2, it omits important causally relevant information: that there are a very large number of other molecular trajectories, compatible with the macroscopic conditions satisfied by the gas (its
temperature, volume, and pressure at t1 and its new temperature at t2) that would lead to the same macroscopic outcome – that is, the new pressure P2. In fact, one may show that for all except a very small set of initial conditions (a set of measure zero) for the
We can express these observations in terms of the ideas about contrastive focus and role of tracing dependency relationships in causal explanation described in Section 1. When one asks for a causal explanation of why the gas is in some macroscopic state – e.g. that of exerting pressure P2, this is most naturally understood as a request for an explanation of why the gas is in that macroscopic state rather than other alternative
macroscopic states –why it has pressure P2 rather than some alternative pressure(s) Pi
different from P2. The microscopic explanation simply doesn’t answer this question, since as stated it tells us nothing about the conditions under which such a macroscopic alternative to P2 would have occurred. By contrast the macroscopic explanation that appeals to the macroscopic state of the gas and the ideal gas law does provide such information, since it allows us to see that if, for example, the new temperature of the gas had been T3 rather than T2, the gas would have evolved to new pressure P3 = nRT3/V. This is not to say that the micro-explanation is entirely unexplanatory – for example, if for some reason we were interested in explaining why the gas ends up at the new equilibrium in the exact molecular configuration G2 rather than the alternative exact molecular configuration G3, the exact configuration of the molecules at their starting point G1 and the details of their subsequent evolution would be highly relevant. But most often, this is not what we are interested in. In the more usual case, where the intended explanandum involves a macro-contrast, the more fine grained and microscopic explanation is not automatically better.
that it answer a what-if-things-had-been-different question or identify conditions such that under changes in those conditions some alternative to the explanandum would be realized and that more generally causal explanation is a matter of tracing dependency relationships and accounting for contrasts. Then assuming that the explanandum in which we are interested is why the gas exerts pressure P2, the micro-explanation leaves out something of importance that is provided by a more macroscopic explanation. In other words, under the right conditions, the interventionist conception favors causal claims and explanations involving more macroscopic variables. This suggests that to the extent that the relationship between some candidate mental cause claim (e.g. that I causes
R in the example above) and the underlying physical/ neural physiological realizations of the candidate cause (N1) and its effect are like the relationship between on the one hand, pressure, volume, and temperature, and, on the other hand, some particular molecular configuration that realizes these variables, the upper level, mental cause claim may be preferable to the claim framed in terms of its neuro-physiological realization N1. More weakly, it seems plainly wrong-headed to think that the microscopic causal claims that appeal to the exact molecular configuration in the case of the gas or to the details of the neuro-physiological realization N1 somehow compete with the more macroscopic causal claims and “exclude” these in the sense of showing them to be false. I will return to this idea below.
3.2) To further explore this point, consider the following example, derived from Yablo (1992). A pigeon has been trained to peck at a target when and only when presented with a red stimulus (that is a stimulus of any shade of red). Suppose that on a series of occasions the pigeon is presented with a stimulus that is a particular shade of scarlet and in each case pecks at the target. Consider the following two causal
claims/causal explanations;
(3.2.2) The presentation of red targets causes the pigeon to peck.
If we adhere to the characterization in M, then both (3. 2.1) and (3.2.2) are true, since in both cases there is an intervention (namely one that changes the color of the target from scarlet to a non-red color) that will change whether the pigeon pecks. Nonetheless, as Yablo argues, there is an obvious sense in which (3. 2.1), like the micro explanation of the behavior of the ideal gas, seems inappropriately specific or insufficiently general, in comparison with (3. 2.2). Or at least, to make a much weaker claim, it does not seem plausible that (3. 2.1) should be judged superior to (3.2.2 ) just on the grounds of its greater specificity. Even if we accept (3.2.1) as a true causal claim, it seems misguided to regard it as in competition with (3.2.2) in the sense that acceptance of the former requires us to regard the latter as false. The basis for these assessments again falls naturally out of the interventionist account of causation and explanation described above. What we are usually interested in when we ask for a causal explanation or cause of the pigeon’s pecking is something that accounts for why it pecks rather than alternatively not pecking at all. There is relevant information about the conditions under which both pecking and not pecking will occur that is conveyed by (3.2.2) but not by (3. 2.1) when both are given their natural default reading. The default reading of 3.1.2.2) with its contrastive focus made explicit is:
(3. 2.2*) The contrast between the targets being red rather than not red causes the contrast between the pigeon’s pecking rather than not pecking
(3. 2.1*) The contrast between the targets being scarlet rather than not scarlet causes the contrast between the pigeon’s pecking rather than not pecking.
In my idiolect, it is most natural to interpret (3.2.1/3.2.1*) as claiming that the pigeon will not peck if the target is not scarlet (but still red). On this interpretation, (3. 2.1) is false. Even if we find this default interpretation uncharitable , it remains true that (3. 2.1) tells us less than we would like to know about the full range of conditions under which the pigeon will peck or not peck5. It is again true that, under the imagined conditions, the presentation of the scarlet target is nomologically sufficient (given the way that the philosophy of mind literature understands the notion of “law” ) for the pigeon to peck, but this just illustrates the point that there seems more to successful explanation or informative casual claims than the provision of nomologically sufficient conditions.
It is also worth noting the obvious point that under a different experimental set-up in which the pigeon was instead trained to peck only in response to the target’s being scarlet, these assessments would be reversed. This again underscores the point that on their most natural interpretation (3. 2.1) and (3. 2.2) have different implications about the manipulability relationships or the patterns of counterfactual dependency that hold in the situation of interest: (3. 2.1) claims that we can manipulate whether the pigeon pecks by changing the target from red to not-red, while (3. 2.2) claims that merely changing whether the target is scarlet will do this. If this is correct, there can be no general preference for (3. 2.2) over (3. 2.1) simply on grounds of its greater specificity – the
5 In part for this reason, I don’t find it useful to worry, as many commentators do, about whether, within a Lewis-style semantics, the “closest” possible world in which the target is not scarlet is one in which it is or is not red. First, the claim that the non-red world is closest seems unmotivated, given Lewis’s official similarity metric. More
appropriateness of each will depend on such facts as whether the pigeon pecks in response to non-scarlet shades of red.
3.3) Finally, suppose (cf. Jackson and Pettit, 2004, p. 172) that John coughs just as the conductor is about to begin his performance and the conductor becomes irritated.
Consider the following two claims
(3.3.1) John’s coughing caused/causally explains the conductor’s becoming irritated
(3.3.2) Someone’s coughing caused/ causally explains the conductor’s becoming irritated.
A general preference for more detailed causal claims/explanations suggests that (3.3.1) is automatically preferable to (3.3.2). Again, however, it is natural to associate these with different contrastive claims. (3.3.1) naturally suggests that the conductor’s irritation is specifically the result of John’s coughing: that the contrast between the actual situation in which John coughs and an alterative in which someone else coughs is
responsible for the conductor’s irritation. (Imagine that the conductor is generally unflappable and undisturbed by coughing but has a particular animus toward John, who he correctly believes is trying to disrupt his performance). That is, the conductor would not have become irritated if anyone else had coughed. By contrast, (3.3.2) does suggest that the conductor would have become irritated as long as there was coughing by anyone. Which of these claims is correct obviously depends on the empirical facts of the
situation – again there are no grounds for supposing that the more specific (3.3.1) is automatically preferable on apriori grounds to the less specific (3.3.2).
requires the truth of the causal claims that figure in those explanations. That is, my contention throughout is that the less specific causal claims are true (if they were not, we could not appeal to them to explain) and that regardless of what we may think about the truth of the more specific claims, they at least don’t exclude the truth of the less specific claims. Thus, on my view, someone who accepts that it is correct to appeal to (3.2.2) to causally explain the pigeon’s pecking cannot at the same time hold that (3.2.2) is false or that its truth is excluded by the truth of (3.2.1).
What do these examples have to do with mental causation? My completely unoriginal suggestion is that claims about mental or psychological causation will be true when the relationship between mental states and their underlying realizations are
relevantly like the relationships between the more macro or less specific causal claims and their underlying, more specific realizations in the examples described above6.
As a simple illustration, consider some research concerning the neural coding of intention to reach carried out by Richard Andersen and colleagues at Caltech (Musallam et al., 2004). These researchers recorded from individual neurons using arrays of
electrodes implanted in the PRR (parietal reach region) of the posterior parietal cortex in
macaque monkeys. Previous research had suggested that this region encodes what Andersen et al. call intentions to reach for specific target – that is higher order plans or goals to reach toward one target or goal rather than another (e.g. an apple at a specific location rather than an orange at some other location) rather than more specific instructions concerning the exact limb trajectory to be followed in reaching toward the target – the latter information being represented elsewhere in motor areas.
Andersen was able to develop a program which systematically related variations in aggregate features of the recorded signals to variations in intentions to reach for specific goals, as revealed in reaching movements and which indeed allowed for accurate forecasting of reaching behavior from these signals. His eventual hope is that paralysed subjects will be able to control the goals toward which prosthetic limbs are directed by forming different intentions, which would then be decoded, the resulting neural signals directing the limb. From a interventionist perspective, this is about as clear a case of mental causation as one could imagine, since the subject uses the formation of one intention rather than another to manipulate the position of the limb.
behavior of individual neurons which is consistent with the holding of the same intention, while some other range of variation in the behavior of individual neurons (associated with a different aggregate firing rate) will be associated with a different intention. In this sense the same intention may be multiply realized in somewhat different patterns of neuronal activity.
Suppose then that on some specific occasion t a monkey forms an intention I1 to reach for a particular goal – call this action R1. Suppose N11 is the particular (token) pattern of firing in the relevant set of neurons that realizes or encodes the intention I1 on this particular occasion. Assume also that there are other token patterns of neural firing,
N12, N13 that realize the same intention I1 on other occasions, so that I1 is multiply realized by N11, N12, etc. The preference for micro or fine grained causation that we are
considering recommends that we should regard NII as the real cause of R1 on occasion t. But this seems wrong for the same reason that it seems wrong to say that it is the scarlet color of the target that causes the pigeon to peck in circumstances in which the pigeon will peck at any red target and wrong to say that it is the specific molecular configuration G1 rather than the fact that the temperature of the gas has been increased to
T2 which is responsible for its new pressure P2. Just as with these two examples, the causal claim/causal explanation that appeals to N11 to explain R1 seems overly specific. It fails to convey a relevant pattern of dependence: that there are some alternatives to N11
accomplish this, since it tells us nothing about the conditions under which alternatives to
R1 would be realized. By way of contrast, appealing to the fact that the monkey’s intention is I1 rather than some alternative intention I2 does accomplish this, assuming (as we have been all along) that there is a stable relationship between the occurrence of I1
(however realized) and R1 and that under I2 some alternative to R1 (reaching toward a different goal) would have occurred.
Note that there is nothing about this argument that relies on the specifically
mental (however this is understood) character of I1 in establishing its explanatory credentials with respect to R1. The argument would proceed in the same way if we instead appealed to neural or physically characterized facts about the aggregate profile – call this A1-- of the firing rates that realize or correspond to I1. In other words, insofar as this aggregate profile A1 corresponds to the different ways N11, N12, N13 of realizing I1, and
A1 leads to R1 and A1 contrasts with whatever aggregate profile of neural activity A2 corresponds to the different intention 12, it will be equally appropriate to cite A1 as causing or figuring in the causal explanation for the monkey’s exhibiting R1.
some possible set of interventions7. Prima facie, at least, (3.4) and (3.5) do not do this. If this appearance is correct, they are not competing causal claims in any sense that requires us to choose between them. Hence, whatever view we take about the relationship between I1 and A1 must be consistent with this fact.
4.
In this section I want to add some clarifications to the argument in section 3 and also to place the argument in a more general context. First, some philosophers8 have argued, with respect to set-ups of the sort under discussion, as follows: (4.1) If the realizer N11 of I1 that actually occurred had not occurred, then some other alternative realizer of I1-- say N12 -- would have occurred instead and would have caused R1. They take this to support the claim that I1 causes (or at least plays some causal role in the occurrence of) R1. This is not the argument made above and I see no reason to accept the counterfactual (4.1)9. The argument that I give above is that we can conclude that I1
7 We should of course distinguish between the question of whether two causal claims are inconsistent and whether thy are different—the latter requires only that they make different claims about what happens under some interventions. It is not clear, however, that (3.4) and (3.5) are different in this sense.
8 See Lepore and Lower, 1987. A somewhat similar argument seems to be suggested in Yablo, 1992, as Bennett (2003) notes, and it may be that Jackson and Petit have
something similar in mind when they speak of a higher level property “programming” for its realizer – if one realizer does not occur, the “program” ensures that another will. The argument may seem particularly natural within Lewis’ framework : it may seem tempting to argue that if N11 actually occurs, then among those worlds in which N11 does not occur, those in which some alternative realizer of I1 occurs are closer to the actual world then those in which the neural realizer of some different intention from I1 occurs. I take no stand on whether Lewis’ theory licenses this sort of inference but a little thought will show that making it leads one into difficulties.
causes R1 because of the truth of various interventionist counterfactuals linking the occurrence of I1 to R1 and the occurrence of alternatives to I1 to alternatives to R1. This argument does not depend upon any claims about what would have happened if N11 had not occurred, although it does of course depend on claims about what would have happened if11 had not occurred.
Next some remarks about the role played by considerations having to do with supervenience and multiple realizability in the arguments just described: In the examples considered in section 3, upper level properties that are causally relevant to other level properties supervene on and are multiply realized by lower level properties. It is
important to realize, however, that what establishes a role for the upper level property— again let’s call it M1 – in the causation of a second upper level property M2 is not just the multiple realizability of M1 and M2 per se (and not even the conjunction of multiple realizability with the existence of laws that in some way link realizers of M1 to realizers of M2) . Instead, what is required is the combination of the right sort of multiple realizability with the existence of a relationship between M1 and M2 such that different values of M1 are systematically associated with different values of M2 and where this relationship is stable or invariant under some range of variations in different lower level realizations of those properties, when these are produced by interventions. In other words, what is required is the existence of a relationship that both involves a dependency between the upper level variables (different values of M1, produced by interventions map into different values of M2) and that is realization independent in the sense that it
continues to stably hold for a range of different realizers of these values of M1 and M2. It
is the presence of this sort of realization independent dependency relationship (hereafter
RIDR) that ensures that interventions that change M1 are stably associated with changes in M2 -- hence that M1 causes M2. 10
To illustrate the significance of this point consider the following example. An ordinary, fair roulette wheel (the operation of which is deterministic at a micro level) is spun by a croupier C. C has a set of possible hand movements Bi for putting the wheel in motion – he can start the wheel at one of a number of positions and he can spin the wheel with more or less force or momentum. Even if the Bi are finely grained from a macroscopic point of view (e.g. they correspond to the maximally fine movements that C can distinguish or control), each Bi will be multiply realized by a range of different exact positions and momenta for the wheel. Similarly, whether the ball ends up in a red or black slot will be multiply realized.
10 The claim that such a RIDR exists is thus importantly different from (and stronger than) the usual claims about “multiple realizability” in the philosophy of mind
literature. Consider the gas law PV=nRT which is RIDR involving temperature,
If the wheel is fair, C will be unable to control or manipulate, by employing one set of hand movements Bj rather than another set Bk, whether the ball will land in a red slot or a black slot—indeed he will be unable to even influence the probability of this happening. For all the different possible Bi, the probability of red will be the same—one half. Consider those occasions on which C pushes the wheel with the specific motion Bk
and on which the ball ends in a red slot R. On each of these occasions Bk and R
presumably will have different micro realizations and, moreover, for each such micro realization of Bk, there will be a law linking it to the micro-realization of R that occurs. (Remember this is a deterministic system). However, within an interventionist
framework, it is not true that
(4.2) Imposing motion Bk on the wheel causes the ball to fall in a red slot. The reason for this is two-fold. First, there are no stable upper level relationships of the form
(4.3) If C pushes the wheel with motion Bi, the ball will land in a red slot.
Instead, when C employs Bi, (for any value of i) whether the ball ends up in a red or black slot depends on the specific micro-realization of Bi that is imposed: generalizations of form (4.3) are not realization independent. Second, there are stable generalizations of the form
(4.4) If C pushes the wheel with motion Bi, the probability the ball will land in a red slot is p.
wheel—what matters is not just the existence of some set of deterministic laws linking the initial conditions of the wheel to the outcome but rather very specific features of these laws and how they relate to the macroscopic predicates used to characterize the behavior of the wheel 11.
In each of the examples considered in Section 3, we in effect assumed that we were dealing with systems that (with respect to the relationships of interest) did not behave like the roulette wheel. For example, in the case of the gas, we assumed both that all (or more strictly virtually all) of the different microstates that realize the same temperature T2 will have the same stable effect on other macroscopic variables like the new pressure measurement P2and that there are alternatives to T2 – e.g., alternative temperature T3 such that all the different micro-realizations of T3 lead to different pressure measurements P3. (In fact, both these claims are roughly true, as an empirical matter, and statistical mechanics gives us some insight into why they are true.) Similarly, in the discussion of Andersen’s research, we assumed that all of the different possible neural realizations of the intention I1 led stably (assuming the right background circumstances) to the same behavior R1, and that the neural realizations of the appropriately different intention I2 would lead to different behavior R2.
As we have seen, discussions of the causal role of mental and other upper level properties have tended to focus on whether it follows just from considerations having to do with multiple realizability (and the absence of type identities ) that such properties are
casually inert. I have argued that this conclusion does not follow. However, undercutting this conclusion certainly does not by itself vindicate claims about the causal efficacy of upper level properties, including mental properties. We also need to ask, in connection with each upper level causal claim, whether the additional requirements embodied in
RIDR are likely to be satisfied. This is a non-trivial empirical question that must be answered on a case by case basis: it is certainly not obvious that the answer to this question is “yes” for many claims of mental causation. I will return briefly to this issue in Section 7 below.
Finally, the general form of the solution described in Section 3 to the problem of how mental properties can play a causal role is not original with me. Broadly similar proposals have been advanced by Yablo (1992) and by Petit and Jackson (2004), among others. Yablo describes his proposal in terms of the requirement that causes fit with or be “proportional” to their effects – that they be just “enough” for their effects, neither omitting too much relevant detail nor containing too much irrelevant detail. In this terminology, I1 fits with (or is proportional to) R1 in a way that N11 does (is) not, since the latter involves too much irrelevant detail. However, Yablo’s treatment relies heavily on essentialist metaphysics in explaining what is involved in a cause being proportional to its effect. I think that this essentialist metaphysics is not necessary and that the intuition behind the requirement of proportionality need only appeal to the considerations invoked in Section 1—an interventionist account of causation, contrastive focus and so on. Roughly speaking, a cause variable will be “proportional” to an effect variable when the pattern of dependence of all alternative possible states of the effect on alternative possible states of the cause is exhibited and there are no additional “irrelevant
distinctions” among alternative states of the cause variable – irrelevant in the sense that these alternatives are not associated with differences in the effect variable12.
Pettit’s and Jackson’s discussion of mental causation is organized around their account of “program explanation” and an associated notion of causal relevance (which they also associate with “instrumental effectiveness” -- see below). A mental state such as the belief that p is causally relevant to some effect A if “variations in how [this mental state is realized remains consistent with invariance in the appearance of the effect [A, of this mental state]” (2004, p. 2.). If we interpret this characterization of what it is for a mental state to cause or be causally relevant to an outcome along interventionist, RIDR lines (that is, that different interventions on the same mental state that involving different realizers of this state lead to the same effect and that the realizations of
different mental states, also produced by interventions, lead to different effects), then the characterization to be essentially the same as (or at least very close to) the
characterization that offered above. However, Jackson and Pettit also distinguish sharply between causal relevance and what they call causal efficacy , where “ a casually
efficacious property with regard to an effect is a property in virtue of whose instantiation, at least in part, the effect occurs” . They associate causal efficacy with the notion of causal “production” and suggest that “relations of causal efficacy” may be “restricted to certain properties of fundamental physics” (2004, p. 61) and perhaps that “causal efficacy is a relation between forces” . (2004, fn. 25, p.61). Causal relevance is thus a broader notion than causal efficacy: casually efficacious properties are causally relevant but a property can be causally relevant or instrumentally effective without being efficacious. According to Jackson and Pettit, this is true of mental states; they are causally relevant to behavior but not in themselves causally efficacious in producing behavior. Instead it is the particular physical realization of the mental state on a given occasion which is causally efficacious in producing behavior. They write:
no matter how the notion of causal efficacy is understood, it is distinct from the notion of instrumental effectiveness. A property will count as instrumentally effective vis-à-vis a particular effect, if it would have been a good tactic for producing the effect to realize that property. But such effectiveness does not entail efficacy: it does not mean that the effect occurred in virtue of the instantiation of the property. (2004, p. 120)
Jackson and Petit suggest that mental states like beliefs do not themselves produce behavior but instead program” for the production of behavior in the sense that they “(non causally) ensure that no matter how it is realized, things will be arranged at the neural and more basic levels , so that behavior is more or less reliably bound to appear”. (2004, p.2)
This distinction between causal efficacy and relevance and the associated idea that explanations that appeal to beliefs provide only information about relevance but not efficacy relations provides a natural opening for critics such as Kim, 1998. Kim claims that a true vindication of the causal status of the mental requires showing that mental states are causally efficacious rather than merely causally relevant; since Pettit and Jackson concede that on their approach mental states are not causally efficacious in producing behavior, they are, according to Kim, really epiphenomenalists about the mental. (Kim, 1998, p. 75)
Many other philosophers similarly distinguish between casual relevance and what they suppose to be a stronger notion of causal efficacy, (although with a different
understanding of the latter notion than Jackson and Petit) and also contend, like Kim, that showing that mental states are causes requires showing that they are casually efficacious in this stronger sense rather than merely causal relevant to their effects . Often (and confusingly13), however, what is meant by “efficacy” seems to amount simply to
